def compare(path,doc):
convert(path,doc)
ann = Annotations(path+doc)
fdoc = folia.Document(file=path+doc+".xml")
#test entities
for ent in ann.get_textbounds():
try:
found=fdoc[ent.id]
text = [str(a) for a in found.wrefs()]
if ent.tail.strip() != " ".join(text):
print "error: not found entity"
print ent
return False
except KeyError:
print "error: not found entity"
print ent
return False
#test relations
for rel in ann.get_relations():
try:
found=fdoc[rel.id]
arefs = found.select(folia.AlignReference)
if not (any(a.id == rel.arg1 for a in arefs) and any(a.id == rel.arg2 for a in arefs)):
print "error: not found relation"
print rel
return False
except KeyError:
print "error: not found relation"
print rel
return False
#test events
for event in ann.get_events():
try:
found=fdoc[event.id]
arefs = found.select(folia.AlignReference)
for role,rid in event.args:
if not any(a.id == rid for a in arefs) :
print "error: not found relation"
print rel
return False
except KeyError:
print "error: not found relation"
print rel
return False
#test attributes
for attr in ann.get_attributes():
try:
found=fdoc[attr.target]
if not any(fattr.cls == str(attr.value) and fattr.subset == attr.type for fattr in found.select(folia.Feature)) :
print "error: not found attr"
print attr
print
return False
except KeyError:
print "error: not found attr"
print rel
return False
print "file "+path+doc+" is OK"
return True
def convert(path,doc):
#path is path to the file without extension
projectconf = ProjectConfiguration(path)
path = path_join(path,doc)
ann = Annotations(path+".ann")
doc = build_text_structure(ann,path+".txt")
add_relations(doc,ann)
add_comments(doc,ann)
#~ ent_set=xml(build_entity_set(doc))
#~ rel_set=xml(build_relations_set(doc))
#~ temp=open ("entiteit_set.xml",'w')
#~ temp.write(ent_set)
#~ temp.close()
#~ rel=open ("relation_set.xml",'w')
#~ rel.write(rel_set)
#~ rel.close()
doc.save(path+".xml")
def convert(data, src):
# Fail early if we don't have a converter
try:
conv_text, conv_ann = CONV_BY_SRC[src]
except KeyError:
raise InvalidSrcFormat
# Note: Due to a lack of refactoring we need to write to disk to read
# annotions, once this is fixed, the below code needs some clean-up
tmp_dir = None
try:
tmp_dir = mkdtemp()
doc_base = path_join(tmp_dir, 'tmp')
with open_textfile(doc_base + '.txt', 'w') as txt_file:
txt_file.write(conv_text(data))
with open(doc_base + '.ann', 'w'):
pass
with Annotations(doc_base) as ann_obj:
for ann in conv_ann(data):
ann_obj.add_annotation(ann)
json_dic = _document_json_dict(doc_base)
# Note: Blank the comments, they rarely do anything good but whine
# about configuration when we use the tool solely for visualisation
# purposes
json_dic['comments'] = []
# Note: This is an ugly hack... we want to ride along with the
# Stanford tokenisation and sentence splits when returning their
# output rather than relying on the ones generated by brat.
if src.startswith('stanford-'):
json_dic['token_offsets'] = stanford_token_offsets(data)
json_dic['sentence_offsets'] = stanford_sentence_offsets(data)
return json_dic
finally:
if tmp_dir is not None:
rmtree(tmp_dir)
def main(args):
# Imports are here so we don't need to wait for them to load unecessarily.
import os
import tkinter
from video import Video
from annotation import Annotations
from state import State
import gui
# Parameters
video_file_path = args.video_file_path
annotation_file_path = args.annotation_file_path
if annotation_file_path is None:
# Expect the following dir structure:
# dataset/
# - videos/
# - annotations/
split_path = os.path.split(video_file_path)
annotation_file_name = split_path[-1].split('.')[0] + '.pkl'
annotation_file_dir = list(split_path[:-1]) + ['..', 'annotations']
annotation_file_dir = os.path.join(*annotation_file_dir)
if not os.path.isdir(annotation_file_dir):
print('Invalid directory structure.')
return
annotation_file_path = os.path.join(annotation_file_dir,
annotation_file_name)
# Load Video
video = Video(video_file_path)
annotations = Annotations(annotation_file_path, video)
state = State(video, annotations)
# Create GUI
gui.App(tkinter.Tk(), state)
# When everything done, release the video capture object
video.close()
generate = True
# "header" and types
stat_types = [("Entities", "int"), ("Relations", "int"), ("Events", "int")]
if options_get_validation(directory) != 'none':
stat_types.append(("Issues", "int"))
if generate:
# Generate the document statistics from scratch
from annotation import JOINED_ANN_FILE_SUFF
log_info('generating statistics for "%s"' % directory)
docstats = []
for docname in base_names:
try:
with Annotations(path_join(directory, docname),
read_only=True) as ann_obj:
tb_count = len([a for a in ann_obj.get_entities()])
rel_count = (len([a for a in ann_obj.get_relations()]) +
len([a for a in ann_obj.get_equivs()]))
event_count = len([a for a in ann_obj.get_events()])
if options_get_validation(directory) == 'none':
docstats.append([tb_count, rel_count, event_count])
else:
# verify and include verification issue count
try:
from projectconfig import ProjectConfiguration
projectconf = ProjectConfiguration(directory)
from verify_annotations import verify_annotation
issues = verify_annotation(ann_obj, projectconf)
issue_count = len(issues)
def get_statistics(directory, base_names, use_cache=True):
# Check if we have a cache of the costly satistics generation
# Also, only use it if no file is newer than the cache itself
cache_file_path = get_stat_cache_by_dir(directory)
try:
cache_mtime = getmtime(cache_file_path)
except OSError as e:
if e.errno == 2:
cache_mtime = -1
else:
raise
try:
if (not isfile(cache_file_path)
# Has config.py been changed?
or getmtime(get_config_py_path()) > cache_mtime
# Any file has changed in the dir since the cache was generated
or any(True for f in listdir(directory)
if (getmtime(path_join(directory, f)) > cache_mtime
# Ignore hidden files
and not f.startswith('.')))
# The configuration is newer than the cache
or getmtime(get_config_path(directory)) > cache_mtime):
generate = True
docstats = []
else:
generate = False
try:
with open(cache_file_path, 'rb') as cache_file:
docstats = pickle_load(cache_file)
if len(docstats) != len(base_names):
Messager.warning(
'Stats cache %s was incomplete; regenerating' %
cache_file_path)
generate = True
docstats = []
except UnpicklingError:
# Corrupt data, re-generate
Messager.warning(
'Stats cache %s was corrupted; regenerating' %
cache_file_path, -1)
generate = True
except EOFError:
# Corrupt data, re-generate
generate = True
except OSError as e:
Messager.warning(
'Failed checking file modification times for stats cache check; regenerating'
)
generate = True
if not use_cache:
generate = True
# "header" and types
stat_types = [("实体", "int"), ("关系", "int"), ("事件", "int")]
if options_get_validation(directory) != 'none':
stat_types.append(("观点", "int"))
stat_types.append(("修改者", "string"))
if generate:
# Generate the document statistics from scratch
from annotation import JOINED_ANN_FILE_SUFF
log_info('generating statistics for "%s"' % directory)
docstats = []
for docname in base_names:
try:
# 在这里获取实体，关系，事件，修改者。
with Annotations(path_join(directory, docname),
read_only=True) as ann_obj:
tb_count = len([a for a in ann_obj.get_entities()])
rel_count = (len([a for a in ann_obj.get_relations()]) +
len([a for a in ann_obj.get_equivs()]))
event_count = len([a for a in ann_obj.get_events()])
try:
user = get_session().get('user')
except KeyError:
user = None
if user is None:
user = '******'
if options_get_validation(directory) == 'none':
docstats.append(
[tb_count, rel_count, event_count, user])
else:
# verify and include verification issue count
try:
from projectconfig import ProjectConfiguration
projectconf = ProjectConfiguration(directory)
from verify_annotations import verify_annotation
issues = verify_annotation(ann_obj, projectconf)
issue_count = len(issues)
except BaseException:
# TODO: error reporting
issue_count = -1
docstats.append([
tb_count, rel_count, event_count, issue_count, user
])
except Exception as e:
log_info('Received "%s" when trying to generate stats' % e)
# Pass exceptions silently, just marking stats missing
docstats.append([-1] * len(stat_types))
try:
user = get_session().get('user')
except KeyError:
user = None
if user is None:
user = '******'
# Cache the statistics
try:
with open(cache_file_path, 'wb') as cache_file:
pickle_dump(docstats, cache_file)
except IOError as e:
Messager.warning(
"Could not write statistics cache file to directory %s: %s" %
(directory, e))
return stat_types, docstats
full_name = temp_paths[1].replace("/", "")
temp = open(app_path + full_name, 'wb')
pickle_dump(sann, temp)
temp.close()
except Exception as e:
Messager.error("Error while caching changes in the annotation file: " +
str(e))
def update_dump(j_dic, file_path):
app_path = WORK_DIR + "/application/"
temp_paths = file_path.split("/data/")
try:
full_name = temp_paths[1].replace("/", "")
temp = open(app_path + full_name, 'wb')
pickle_dump(j_dic, temp)
temp.close()
except Exception as e:
Messager.error("Error while caching changes in the annotation file: " +
str(e))
if __name__ == '__main__':
millis = int(round(time.time() * 1000))
print millis
ann = Annotations("/home/sander/Documents/Masterproef/brat/data/test")
sann = SimpleAnnotations(ann)
print filter_folia(sann)
millis = int(round(time.time() * 1000)) - millis
print millis
temp.append(i)
ids.add(i)
if isinstance(temp_ann, BinaryRelationAnnotation):
for i in temp_ann.get_deps()[1]:
if not i in ids:
temp.append(i)
ids.add(i)
for i in temp:
temp_ann = ann.get_ann_by_id(i)
if isinstance(temp_ann, TextBoundAnnotation):
recursive_ann(temp_ann, con, ids, ann)
if __name__ == "__main__":
from annotation import TextBoundAnnotation, TextAnnotations, EventAnnotation, BinaryRelationAnnotation
proj = ProjectConfiguration(
"/home/sander/Documents/Masterproef/brat/data/brat_vb/sentiment")
ann = Annotations(
"/home/sander/Documents/Masterproef/brat/data/brat_vb/sentiment/sporza"
)
#SPEED TEST
import time
millis = int(round(time.time() * 1000))
print millis
vrules = ValidationRules(proj)
for i in vrules.validate(ann)[0]:
print str(i)
millis = int(round(time.time() * 1000)) - millis
print millis
